Monitoring of Diaphorina citri Kuwayama (Hemiptera: Liviidae) populations is an important component of efforts to reduce damage caused by huanglongbing, a devastating disease it vectors in citrus groves. Currently, D. citri is monitored primarily by unbaited sticky traps or visual inspection of trees. A potentially more effective method might result from attracting males to vibrational communications produced by females. Males call with wing-buzzing substrate-borne vibrations while searching for females on tree branches and stems. When nearby receptive females detect the calls, they reply immediately in synchronized duets that help direct the males towards them. The spectral and temporal patterns of the duets have been analyzed in previous studies and have been mimicked successfully with computer-operated vibration exciters. Males and females both respond to signals produced by either sex but display different behaviors during duets. To devise practical methods to attract and trap males with vibrational signals in field environments, a microcontroller platform was tested for capability to control inexpensive vibration sensing and output devices. The microcontroller was programmed to send mimics of different D. citri signals to a piezo buzzer for substrateborne broadcast. A mimic that elicited strong female responses was tested in bioassays that jointly compared it with other previously bioassayed signals, and the response to the mimic was found to be statistically comparable to that elicited by a recorded male call. The successful result suggests there is opportunity to develop microcontroller systems further as a means of trapping psyllids.
The need to develop improved methods to detect and suppress populations of Diaphorina citri Kuwayama (Hemiptera: Liviidae) (Sétamou et al. 2008; Wenninger et al. 2009b; Hall et al. 2012; Grafton-Cardwell et al. 2013), an economically important vector of huanglongbing (Bové 2006; USDA National Invasive Species Information Center 2013; Browning 2013), has stimulated studies to understand and co-opt the vibrational communication and mating behaviors of this pest for detection and trapping (Wenninger et al. 2009a; Rohde et al. 2013). Mate-seeking males produce substrate-vibration calls intermittently while exploring along branches of a host citrus tree. When a receptive female replies back in a synchronous duet, the male performs a directed search to locate her, continuing to call occasionally during the searching process and using her replies to help target his search. Males and females produce vibrational signals with similar spectral and temporal characteristics (Wenninger et al. 2009a). Potentially, a trap that attracts male D. citri effectively with vibrations might enable more accurate sampling to better determine when populations begin to increase and treatment is needed (e.g., Stansly et al. 2009).
In a recent proof-of-principle study (Rohde et al. 2013), a laptop-controlled vibration exciter produced D. citri signal mimics that attracted males searching on small citrus trees in a laboratory environment. Males respond to signals recorded from either female or male conspecifics, particularly those that contain multiple harmonics of about 200 Hz at frequencies between 1000-2000 Hz. Signals derived from female replies, male calls, and computer program outputs have been used successfully to elicit both male attraction and female replies. Preliminary investigations using the laptop-exciter system suggest that males engage in search behaviors more frequently when the mimics are synchronized to follow immediately after a call than when they are presented at random or are absent.
The demonstration that synchronized mimics of duetting signals are attractive to male D. citri in a laboratory setting suggests that such methods might lure mate-seeking males to traps in field environments as well. However, much of the technology associated with detection, analysis, and production of insect vibrational communication signals is difficult or costly to employ in the field (Cocroft & Rodriguez 2005; Mankin et al. 2010, 2011). To transfer this technology into effective trapping systems for D. citri males in citrus groves, it would be preferable to substitute lowpower, low-cost, compact devices for the laboratory instrumentation. Here, we describe tests of the capability of a battery-powered, low-cost microcontroller platform operating an inexpensive piezoelectric buzzer to produce effective mimics of D. citri communicatory vibrations.
A major objective in programming the microcontroller output was to produce multiple harmonics of 200 Hz between 1000 and 2000 Hz, as was observed previously in signals that successfully elicited male attraction (Rohde et al. 2013). In addition, the harmonics usually were 5-10% higher near the middle of the signals than at the beginnings and endings, so this signal characteristic also was incorporated into the mimics. To determine if the output had relevance to D. citri behavior, the mimics were tested in bioassays that compared proportional responses of females to mimics with responses to previously bioassayed D. citri signals, where proportional response was measured as the fraction of signals presented that elicited a female reply.
Materials and Methods
Insects and Experimental Arena
Nymphs obtained from a rearing colony maintained at USDA-ARS-CMAVE, originally collected from citrus in fields near Ft. Pierce, Florida (Hall et al. 2007), were placed individually into isolation chambers, each prepared in advance by filling a 21-cm-long, 3.75-cm-diam cone-tainer (model SC10, Steuwe and Sons, Inc., Tangent, Oregon) with potting soil and adding a small citrus seedling. The isolation chambers were capped with 10-cm-long tubing, screened at the top, with four additional screened holes spaced ca. 1 cm below the top (see additional details in Paris et al. 2013). Soil moisture was maintained by placing the isolation chambers in racks over water-filled trays. The nymphs, kept on a 16:8 light cycle at 25-30 °C, emerged as adults within 2-6 days and females were selected for bioassays 3-9 days after emergence. Bioassayed females were kept in their original isolation chambers, separate from other conspecifics, and remained virgin until testing. They were transferred to the CMAVE rearing colony after the tests.
Bioassays were performed inside a vibrationshielded anechoic chamber, 4-10 h after the beginning of photophase, on a 25-cm-height Duncan grapefruit tree (Citrus paradisi Macf.; Rutaceae) or a 30-cm-height Hamlin sweet orange tree (C. sinensis L. Osbeck; Rutaceae). Light was supplied by a three 60-W floodlamps from ca. 1 m above the tree. The bioassays were observed remotely using a videocamera (model HDR-SR1, Sony Corp. New York, New York), 0.5 m to the side, that fed its signal to a monitor outside the chamber.
Signals produced in bioassays were detected by an accelerometer (model 4371, Brüel and Kjær [B&K], Naerum, Denmark) that had been attached by an alligator clip near the base of the tree. The signals were fed from the accelerometer through a charge amplifier (model 2635, B&K) to a signal analysis system (model 4300B, Kay Elemetrics Corp, Lincoln Park, New Jersey) outside the anechoic chamber for digitization, recording, and real-time spectral analysis. Additional temporal pattern and spectral analyses of digitized signals were performed using custom-written (Matlab R2011b, MathWorks, Natick, Massachusetts), Raven (Charif et al. 2008), or Audacity ( http://audacity.sourceforge.net) software.
Microcontroller-Buzzer and Laptop-Vibration-Exciter Systems for D. citri Mimic Broadcasts
The feasibility of low-cost production of D. citri vibrational signal mimics was considered by connecting an inexpensive, open-source microcontroller prototyping platform (Arduino Uno, Arduino Inc., Italy http://arduino.cc/en/Main/ArduinoBoardUno) to a circuit containing a 6-12 V-DC power source, an on-off switch, and a piezoelectric buzzer (Fig. 1). The microcontroller platform is compact (8 by 5.4 by 3.2 cm) and fits easily into many commonly used insect traps. A standard 8-AA-cell, 6.0 by 3.0 by 6.2-cm battery holder can fit inside the trap if it is not convenient to attach it to a nearby branch or trunk, and a 9-V battery can be used for short-duration applications. Several preliminary laboratory tests were conducted by operating the system with 9-V bat- teries, and preliminary laboratory and field tests were conducted by operating the system with the AA-cell packs to confirm estimates (see http://arduino.cc/en/Main/ArduinoBoardUno) that, at a typical rate of current usage of 50 ma, the 8 AA cells could provide power for periods up to nearly 2 days and a 9-V battery for approximately 6 h. In these cases, the batteries first were connected to the Arduino system to provide continuous power, and the capability of the system to power the buzzer was tested multiple times each hour towards the end of the expected operating period.
Initially, the microcontroller was programmed to oscillate 6 of the 14 digital output pins at frequencies of 200, 400, 600, 800, 1,000, and 1,200 Hz. These 6 frequencies then were combined by adding the signals from the 200-Hz pin to those from one or more of the other output pins to the piezoelectric buzzer. Resistance (e.g., 10 k?? in Fig. 1) was added to the circuit as needed to reduce or increase signal output to biologically relevant levels (see next section). After the initial tests of system operation, the programming of the output pins was modified slightly to produce a mimic with multiple harmonics that increased and then decreased in frequency during the course of a signal, as had been observed with recorded calls and replies in Wenninger et al. (2009) and Rohde et al. (2013) (see RESULTS). The resultant mimic, denoted as mop2_12, was evaluated by attaching the buzzer to a side branch of the tree with an alligator clip and monitoring the responses of females to the microcontroller-buzzer system broadcasts.
For consistency, a 9-V, 17-mm-diam buzzer (9S3164, Taiyo Yuden, Tokyo, Japan) with a resonant frequency of 7.6 kHz was used in all the bioassays and most of the preliminary tests. A variety of inexpensive buzzers of different sizes and resonant frequencies are available from different manufacturers, however, and additional tests were conducted to consider whether differences among buzzers significantly affected characteristics of signals transmitted through the tree. Three other buzzers with different resonant frequencies were tested, manufactured by PUI Audio Inc., Dayton Ohio: 2.0 kHz (No. AB2720B), 3.6 kHz (AB2036B), and 4.6 kHz (AB2746B). Each buzzer was connected to the Arduino system and then attached to a side branch of the tree. The buzzer signals were monitored by the accelerometer and their amplitudes and frequency spectra were analyzed using Audacity software.
Bioassays of Female Response to Microcontroller- Buzzer and Laptop-Vibration-Exciter Signals
The responses of females to the buzzer broadcasts were placed in context by conducting bioassays in which the females were presented with mop2_12 mimics as well as 4 conventionally produced signals that had elicited a wide range of high to low rates of proportional responses: a recorded male call, a recorded female reply, a synthetic mimic (mh1200), and white noise (Rohde et al. 2013). The conventionally produced signals were played by a laptop computer using QuickTime (Apple, Inc., Cupertino California) to a vibration exciter (Model 4810, B&K) attached with a small push rod near the middle of the main stem of the tree.
In each bioassay, a female was coaxed from a small vial onto a leaf near the top the tree. After she settled on the flush and began feeding, the observers left the anechoic chamber to avoid further disturbance. Thirty examples each of the mop2_12-mimic broadcasts and the 4 signals from the vibration exciter were triggered in random order (generated by the Matlab function, randperm) at intervals of 5-10 s. The responses to each signal were noted. As in previous bioassays, the proportional response to each test signal was measured as the fraction of playbacks to which the female replied. Eleven females were bioassayed.
In preparing for the bioassays, the amplitudes of signals from the microcontroller-buzzer and laptop-vibration-exciter systems were adjusted to approximate the amplitudes of replies from several females in preliminary tests. The mop2_12-mimic amplitude was adjusted by increasing or decreasing the circuit resistance, which typically was set at 10 k??. The signals from the vibration-exciter system were adjusted by changing the amplitude of the output from the laptop computer. These settings were used for subsequent bioassays, with occasional adjustments when signal levels of female replies and test signals failed to remain similar in amplitude.
Statistical Analyses
Significance of differences among responses to different signals was confirmed by applying a nonparametric one-way repeated measures Friedman's test (SAS Institute 2004). Nonparametric Wilcoxon signed-ranked tests (SAS Institute 2004) then were applied to compare separately the proportional responses to different signals.
Results
Microcontroller-Buzzer Broadcast Characteristics
An example of a series of signals produced by combinations of output from different pins of the microcontroller platform to the 7.6-kHz piezo buzzer is shown in Fig. 2. The 200 Hz pin operated during each of the series, and multiple har monics of the 200 Hz fundamental appear clearly. Contributions from operation of the 400, 600, 800, 1,000, and 1,200 Hz pins, respectively, are seen in the dashed boxes. In this example, the output pins were switched manually, and broadband background noise was present during the period when the operator switched the signal from one pin to another.
Fig. 2.
Spectrogram of a series of signals transmitted to tree by microcontroller output to piezoelectric buzzer. All signals in this series contained output from 200-Hz pin. Solid bars with pin labels below the bottom of the spectrogram mark the times when signals from 400-, 600-, 800-, 1000- or 1200-Hz pins were added to the 200-Hz pin signal. Dotted boxes indicate frequencies where contributions from the added pin were observed. Harmonics of the initial, fundamental frequencies also appear in each signal of the series. Darker lines in the spectrogram indicate relatively higher signal energies; for example, the line near 1 kHz contributed from the 1000-Hz pin, the line near 1.2 kHz contributed from the 1200-Hz pin, and the lines near 1.4 and 1.8 kHz which appeared whenever the 200-Hz pin was turned on.
The combination of 200 and 1,200 Hz output pins (last signal in the series of Fig. 2) was one of several in the series that produced multiple harmonics with strong energy between 1 and 2 kHz, which had been found previously to be important for eliciting high proportions of female replies (Rhode et al. 2003). This signal was selected as the first candidate for bioassay testing. To modify the output so that the frequencies of harmonics increased and then decreased during the course of the signal, the programming of the Arduino output pins was adjusted so that the 200 Hz and 1,200 Hz fundamental frequencies were increased and then decreased in steps of 0.833 and 5 Hz, respectively, at intervals shown in Table 1. An example of the resultant signal (mop2_12-mimic) and a female reply is shown in Fig. 3, along with examples of other test signals and replies.
The spectra of signals transmitted from buzzers with 2.0-, 3.6-, and 4.6-kHz resonant frequencies to the accelerometer at the base of the tree are shown in Fig. 4. Because each buzzer was driven by the same 200 and 1200 Hz signals from the Arduino microcontroller, it was expected that peak frequencies would occur at multiples of 200 Hz harmonics, but the relative amplitudes might be different. In this case, the 2 kHz buzzer produced the highest signal levels across most of the spectral range, but all of the buzzers produced multiple harmonic peaks between 1–2 kHz.
Table 1.
Frequencies and timing of tones programmed to microcontroller output pins for rising and falling phases of Diaphorina citri signal mimics.
Female D. citri Proportional Responses
Females replied to a significantly greater fraction of male-call and mop2_12-mimic signals than to female-reply and mh1200-mimic signals, and all of these responses were greater than to white noise signals (Table 2). All but 3 females exhibited a greater proportional response to the male-call signal than to any other. Two of these exhibited a greater proportional response to the mop2_12-mimic signal and one to the female-reply signal. Three females exhibited 0.9 or greater proportional response to the male-call signal but exhibited less than a 0.3 proportional response to any of the other signals. This suggests that D. citri females exhibit a range of selectivity to different forms of male calls, as has been observed also with other psyllids (Percy et al. 2006).
Discussion
This study explored the possibility of constructing an inexpensive, low-power device capable of producing synchronized duetting vibrations serving as a lure for trapping male D. citri searching for females in the branches of citrus tree canopies. The results above indicate that readily available microcontroller systems can be programmed easily to produce signals with multiple harmonics, confirmed to be of behavioral relevance to male and female D. citri. There is evidence also, see last section below, that such systems have satisfactory power usage for short to mid-term trapping durations.
Importance of Multiple Harmonics to Attractiveness of D. citri Communication Mimics
The 4 calls and mimics in Table 2 and Fig. 3 were similar in that all of them contained energy at multiple harmonics of 200 Hz between 1-2 kHz, although there many differences in other acoustic characteristics. They all elicited significantly greater behavioral responses than white noise that contained a random distribution of all the stimulatory frequencies. This result is supportive of a hypothesis that the presence of energy at multiple harmonics of about 200 Hz between 1-2 kHz may be important for eliciting female D. citri responses, as observed in a previous study (Rohde et al 2013), while differences in relative amplitudes of harmonics and differences in temporal patterns may have lesser effect. Males also may focus on the presence of multiple harmonics rather than specific frequencies or temporal patterns, considering that in Fig. 3, female replies exhibited both spectral and temporal pattern variability. In addition, we observed multiple occurrences of a female beginning her reply before a male call had ended, which suggests that the total duration of the male call is not a significant determinant in a female's decision to produce a reply.
Part of the reason why D. citri may exhibit some flexibility in the relative amplitudes of different harmonics in their duets is that these aspects of the vibrational signals can be modified in unpredictable ways as the signals are transmitted between duetting partners along the plant substrate (Cocroft et al. 2006; Hambric 2006; Mankin et al. 2008, 2011). In Figs. 2 and 4 for example, only 2 frequencies are outputted to the piezoelectric buzzers, but multiple harmonics of different relative amplitudes can be detected by the accelerometer at the base of the tree. The amplitudes of the harmonics could be different if the buzzer were attached to a different branch or a differently sized tree (Hambric 2006). Only the presence of energy at multiple harmonics, but not a specific pattern of relative amplitudes or durations was found consistently in successful signals observed in Wenninger et al. (2009a), Rohde et al. (2013), and this report.
Fig. 3.
Oscillograms and spectrograms of microcontroller-buzzer or laptop-vibration-exciter signals of different spectral and temporal patterns and the associated replies from test females: A) male-call, B) mop2_12-mimic, C) female-reply, D) mh1200-mimic. White noise is not shown in this figure, but an example of white noise can be seen in Fig. 3 of Rohde et al. (2013). All oscillograms were set at the same vertical scale relative to the signal received at the accelerometer (relative amplitude), and areas of darker shading in the spectrograms indicate higher relative energies at those frequencies and times.
D. citri Vibration Trap Power Usage and Cost Analysis
Power usage is a practical concern for any electronic system in field environments away from an electrical grid. The Arduino system is specified to draw about 50 ma current under routine conditions (see http://arduino.cc/en/Main/ArduinoBoardUno). At this rate, sets of 8 AA cells could power an intermittently operating system for 2 days. A 9-V battery could power the system for about 6 h. These approximate battery lifetimes were confirmed in several preliminary studies (see METHODS).
According to the instruction manual for the microcontroller chip, reductions in power usage could be achieved by placing parts of the system in sleep mode when not in use. Indeed, recent preliminary tests indicate that the Atmega328 microcontroller chip (Atmel Corp, San Jose, California) on the Arduino Uno system not only has the capability to spectrally modulate the signal output, as with the mop2_12 mimic, but has sufficient capability to perform other tasks concurrently with signal production, such as operation of a light sensor to set the system into sleep mode during darkness. In an initial test of this capability, it has been possible to operate a system for 4 days before loss of power. Also, careful selection of piezo buzzers may increase the effective range of a trap for a given level of power usage. The results in Fig. 4 suggest that, for a given power output, signals from 2-kHz buzzers might be detectable over longer distances than the signals from the higher frequency buzzers because the resonant frequency is closer to the most behaviorally relevant signals.
Table 2.
Means ± SE of female Diaphorina citri proportional responses to microcontrollerbuzzer or laptop-vibration-exciter produced signals of different spectral and temporal patterns, arranged in order of high-to-low response.
The $40-$60 cost of these microcontroller-sensor systems is greater than the cost of sticky traps but there are potential benefits from improved trapping efficiency and reusability. Certainly the cost is orders of magnitude lower than the cost of the laboratory equipment it replaces. In addition, experience with programming and operation of these systems indicates that sensors to detect vibration, temperature, humidity, and light could be added and controlled at reasonable cost, which could make such devices even more useful as entomological research tools.
Acknowledgments
We thank Betty Weaver, Jane Sharp, and Everett Foreman (USDA ARS CMAVE) for insect rearing, bioassay testing, and signal analysis assistance. Funding was provided in part by a grant from the Citrus Research and Development Foundation. The use of trade, firm, or corporation names in this publication is for the information and convenience of the reader. Such use does not constitute an official endorsement or approval by the USDA of any product or service to the exclusion of others that may be suitable. The USDA is an equal opportunity provider and employer.
